[LMH]FPGA / microcode

[Hans Hübner]

On Fri, 5 Mar 2004, Eric Blossom wrote:

> Another thing to consider is the basic "back of the envelope
> calculation".   For the sake of argument, assume that you're using a
> Spartan 2E or Spartan 3 (The spartan 3 has bunches of embedded
> multipliers.  [...]

Speed is not the (major) point which drives the wish for an Explorer emulator.
The original Explorer was indeed fast enough for a number of serious tasks, so
having an implementation of it which is, say, ten times faster than the
original hardware gives a lot of speedup for new applications.  Price is much
more influencial.  If I read the latest comments correctly, a FPGA explorer
could cost around $100 in chips, which is really in the (grown-up) toy price
range.

The reason to want Lisp hardware is the fact that such a CPU would eliminate a
lot of complexity levels which are present when using LISP on a general
purpose machine.  Lisp hardware puts the whole computer, and not only a
virtual image of a non-existing system, directly under control of the LISP
code you work with in your development environment.  If you need to work at
high levels of abstraction, you can formulate suitable algorithms readably.
If you need to handle interrupts within a certain amount of time, you can
again formulate solutions in the same language and environment which is your
high-level problem solving tool.

Using the Explorer as starting hardware has the beauty that a lot of work has
already been done by smart people.  For example, TI's Emacs implementation
Zmacs seems to be very good - at least from what I read from the excellent
documentation.  Having a working development environment within the new
plattform is a really big plus, since otherwise you almost certainly end up
cross compiling from Windows or Unix and having to spend time fixing the
broken development environment instead of the target platform.

It is true that the limited word length of the Explorer restricts the
usefulness of any new implementation to problems of a certain maximum size.
Still, I think it would be a excellent platform for applications like mail and
personal information management as well as text processing.  To me, these
applications are not solved satisfactory by current software environments and
I'd rather store and process all my email, contact information and
appointments within a portable solid-state Lisp system than in Outlook or Unix
files.

There have been a lot of ambitious open source developments for advanced
operating systems and development environments.  None of the projects I am
aware of have had any significant visibility, unless they have been directly
funded and influenced by the industry.  Using the existing Explorer code base
would have such a funding in form of the intellectual property of TI (or
whoever), which truly pushes the whole effort into some uncertainity.  I am
very curious to see this progress and I hope that enough free energy can be
accumulated to get a new Explorer implementation running.

Best regards,
Hans